[ BENCHMARKS ] measured results

source-linked numbers.

Each result set links to its source. The landing page shares the same data, so summary and detail stay aligned.

2.35x

TSBS geomean vs reference

3.57x

master suite overall geomean

3.16x

DBB total query speedup

3.08x

DBB per-query geomean

[TSBS] --- result set

TSBS mean query time

Mean query time on the same time-series workload. Lower is better; l normalized to 1.00x.

Source on GitHub

l 1.00x

QuestDB 5x

TimescaleDB 30x

InfluxDB 64x

ClickHouse 195x

[MASTER] --- result set

master-benchmark cross-runtime suite

A 125-op comparable subset of the 135-op cross-runtime suite, grouped by primitive family and reported as section geomeans.

Source on GitHub

scoc 6 9.38x

cmp 10 6.58x

coc 15 5.41x

reduce 12 4.49x

where 7 3.55x

scan 4 3.23x

arith 15 3.09x

la 4 3.09x

hdb 12 2.80x

temporal 13 2.60x

cmpd 15 2.54x

sort 8 2.36x

map 4 2.10x

overall 125 3.57x

[DBB] --- result set

db-benchmark 50 GB groupby

The L solution for the db-benchmark groupby task against DuckDB on a 1e9-row CSV, with row counts and checksums validated.

Source on GitHub

load: l CSV read + partitioned store

73.5 s

load: DuckDB CSV into memory

27.1 s

total query time

l 56.6 s vs DuckDB 178.8 s

total query speedup

3.16x

per-query geomean

3.08x

validation

all 10 queries validated

q	query	l -s	l s	DuckDB s	DuckDB/l
q1	sum v1 by id1	0	0.103	0.262	2.54x
q2	sum v1 by id1,id2	48	0.495	0.505	1.02x
q3	sum v1, mean v3 by id3	16	1.599	34.045	21.29x
q4	mean v1,v2,v3 by id4	0	0.128	0.153	1.20x
q5	sum v1,v2,v3 by id6	16	3.804	26.060	6.85x
q6	median v3, sd v3 by id4,id5	16	1.679	3.941	2.35x
q7	max v1 - min v2 by id3	16	1.392	15.337	11.02x
q8	top-2 v3 by id6	16	8.628	24.715	2.86x
q9	corr(v1,v2)^2 by id2,id4	48	0.542	0.651	1.20x
q10	sum v3, count by id1..id6	16	38.270	73.165	1.91x